Horizontal and vertical interactions in local fiscal slack: a spatial analysis of local governments in New York state

ABSTRACTWith the recurring economic crises, the literature on fiscal savings behaviour has expanded its scope from state governments relying on budget stabilisation or rain-day funds to local governments mainly using general fund balances. While many studies have explored the determinants of local savings behaviour, few have taken fiscal interactions among local governments into account. To fill out this gap, this study explores whether spatial interdependence in fiscal slack exists horizontally between counties and vertically across counties and municipalities through the unreserved fund balance of local governments in New York State from 1996 to 2016. The empirical findings indicate that horizontal interdependence in fiscal slack among county governments exists in the short-run, while vertical interdependence is observed across counties and municipalities in the long-run.KEYWORDS: local fiscal slackunreserved fund balancefiscal interdependencehorizontal and vertical fiscal interactionsspatial analysis Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1. In particular, the Antideficiency Act Amendments of 1982 (ADA) has triggered government shutdowns and imposed institutional restrictions on the federal government’s ability to stabilise the economy.2. Interchangeably used with fiscal savings and slack resources, the term fiscal slack generally refers to the pool of resources available to a government in excess of what is necessary to produce a minimum level of services (Jacob and Hendrick Citation2012, 17). Fiscal slack is usually accumulated in the forms of unreserved fund balance at the local level and budget stabilisation funds or rainy-day funds at the state level.3. In this study, we express spatial effects in local fiscal slack using such neutral terms as fiscal interactions and interdependence as our study is not designed to identify potential sources of the interrelationships (e.g., spillovers, yardstick competition, mobility-based competition). Instead, we theoretically infer the mechanisms of fiscal interdependence using concepts such as spillovers, mobility-based or yardstick competition, and policy learning or emulation in the theoretical background section.4. The Tiebout hypothesis, commonly recognised as the origin of the mobility-based competition model, assumes that residents have perfect mobility and information on each local government’s tax rates and spending levels; thus, residents have incentives to choose and migrate to local governments that provide their preferred fiscal packages (Tiebout Citation1956). As a result, local governments come under pressure to compete with other local governments to prevent such mobile resources that provide tax bases from leaving and attract new ones.5. Among the few, Guo and Wang (Citation2017) found that a Florida county government’s decision on an unreserved general fund balance level is affected by not only its own millage rate and reliance on intergovernmental revenues but also its neighbouring counties’. Another one is Shon and Kwak (Citation2020), which found that the fund balance of a county in California becomes lower with an increase in the fund balance of its neighbouring counties.6. They are New York, Kings, Bronx, Richmond, and Queens. Unlike the other 57 counties, the administrative powers of these five boroughs (parallel with counties) are severely limited, even weaker than those of the NYC, because the NYC governs all aspects in those five boroughs. Owing to this difference in fiscal power from the other counties, the empirical analysis excludes the 5 boroughs within the NYC. Compared with the other sub-counties, such as 62 cities, 933 towns, and 535 villages as of 2019, the NYC has the much greater powers; hence, it is excluded as an outlier from the sample.7. The office of the New York State Comptroller (Citation2019a) provides financial information for only 923 towns, excluding five coterminous towns-villages, and the financial information data includes only 560 villages because some villages were merged into other jurisdictions.8. The unreserved category is composed of designated and undesignated fund balance. Local governments have more discretionary powers in using the undesignated one, and they can change the purpose of the designated one with ease.9. The GASB-54 has required local governments to follow the five fund balance categories (Governmental Accounting Standards Board Citation2010). (a) A nonspendable fund balance is one that cannot be spent because of legal and contractual requirements, and is not converted to cash. (b) A restricted fund balance should be used when constraints placed on the use of resources are imposed by external parties, constitutional provisions, or enabling legislation. (c) A committed fund balance includes amounts that the highest level of decision-making authority in a county commits for a specific purpose. (d) An assigned fund balance includes amounts that a county intends to expend for specific purposes. However, this balance is neither restricted nor committed, and should be reported as an assigned fund balance, except for stabilisation arrangements. (e) An unassigned fund balance is a residual classification of amounts that are not included in the first four classifications in the general fund. Because local governments follow the GASB rules and report their financial conditions in the standardised format of the New York state, any potential differences in that local governments provide services and collect revenues do not matter for the empirical analysis.10. The equation for the linear trend estimate is logTRit=α+β1i+β2t+εit, where TRit denotes the tax revenue of a county government i in a fiscal year t.11. The results include three elections for President, Governor, and U.S. Congress. When an election did not happen, the values were coded with the previous election results (Adams et al. Citation2004; Levitt Citation1996).12. After the estimation of SDM, we performed the first test whether θ=0 and ρ≠0, which leads to that SAR is more appropriate. Then, the second test was performed whether θ=−βρ), which indicates that SEM is more appropriate. However, the test results rejected the null hypotheses, and revealed that θ≠0 and ρ≠0 (with a χ2=51.13; p < 0.001), as well as θ≠−βρ (with a χ2=49.16; p < 0.001). In addition to the LM-test, the autocorrelation (AR1) test, the panel (groupwise) heteroscedasticity test, and multicollinearity test were conducted. The test results reported that the AR(1) and heteroscedasticity exist; thus, the dynamic estimates were obtained through the Huber-White standard errors and using the robust estimator of variance in order to remove any potential endogeneity. However, no multicollinearity was observed (Mean VIF = 8.78).13. The Hausman test also rejected the null hypothesis, with a χ2 test statistic equal to 75.71 (p < 0.001).14. The spatial matrix was contiguity-based, and then row-standardised, which can obtain the means of the spatially lagged variables for the empirical analysis. There was no systematic difference between rook and queen spatial-weight in the counties, NY, except for New York city.15. Apart from the dynamic estimates in Table 2, the static estimates are also provided in Table A1, Appendix for the purpose of providing information to readers.Additional informationFundingThis work was supported by the Ministry of Education of the Republic of Korea and the National Research Foundation of Korea (NRF-2022S1A5A8052255). It was also supported by the Hankuk University of Foreign Studies Research Fund.